Research On Prediction And Optimization Of Drilling Rate Based On Deep Learning

Drilling rate is an important index in drilling engineering,which directly affects the efficiency of drilling engineering.Therefore,the prediction and optimization of drilling rate has always been an important research direction in the industry.Limited by many factors,there are still many problems to be solved.Aiming at the practical problems of deepwater drilling engineering,this dissertation proposes a standardized drilling data processing scheme based on deep learning,including data preprocessing,drilling rate prediction and optimization.The main work is as follows:Firstly,aiming at the problem of outlier detection in drilling data preprocessing,the performance of outlier detection based on Isolation Forest,Elliptic Envelope,Local Outlier Factor and GNN model is analyzed,and a fusion outlier detection method based on Isolation Forest and GNN is proposed.The detection performance of each method is evaluated using autoencoder and reconstruction error,which verifies the effectiveness of the fusion outlier detection method.Secondly,aiming at the accuracy of drilling rate prediction,the basic characteristics of deepwater drilling big data are studied,and the prediction performance of multiple linear regression,RNN model,LSTM model and TCN model in deepwater drilling data is analyzed.The Feature Analysis and Time-series Memory Network applicable to drilling rate prediction are proposed,which verifies the prediction performance of the proposed model and lays the foundation for deepwater drilling rate optimization.Finally,aiming at the problem of real-time optimization of deepwater drilling rate,the drilling rate optimization process is designed and the performance is evaluated by different methods.The goal of model adjustment in the drilling rate optimization task is ranking,and the Feature Cross and Time-series Memory Network suitable for the drilling rate optimization process is proposed,which realized the application of artificial intelligence method in deepwater drilling rate optimization and the effectiveness of the combined drilling rate prediction and optimization scheme is verified.